Losses of potassium (K) are inevitable in grazed New Zealand dairy pasture. Losses result from the removal of K in animal products, transfer of excreta to the dairy shed and raceways, and K teaching below the effective plant rooting zone. In this study the fate of K in a grazed dairy pasture was investigated in order to quantify the extent of these losses more precisely. As a consequence an improved methodology for predicting K fertiliser requirements has been developed. A field experiment on the Tokomaru silt loam soil (yellow-grey earth) showed that immediately after a urination event in July up to 59% of the urine K moved preferentially through a network of soil macropores to below a soil depth of 15 cm. The remaining 41% of the urine K was retained in this depth of soil in the form of exchangeable soil K+. Over the following spring and summer, plants were able to recover the urine K retained in the 0-15 cm depth of soil. The field experiment also showed that pasture plant roots growing in urine affected soil derived their K requirements mainly from the 0-15 cm depth of soil. Any K that moved beyond a depth of 15 cm due to leaching can be considered to be a loss of K from the grazed dairy pasture. In a laboratory study, two soil types of contrasting mineralogy (the Tokomaru silt loam, a yellow-grey earth, and the Egmont brown loam, a yellow-brown loam) were incubated with either dairy cow urine potassium chloride. Within 24 hours the soil pH in the urine treatment increased by one unit on both soil types. This increased soil pH was evident for 41 days in the Tokomaru soil and for the duration of the experiment (106 days) in the Egmont soil. For the Egmont soil, which contains larger amounts of pH dependent negative surface charge, this increase in soil pH resulted in greater K+ adsorption by the soil from the urine treatment compared with the potassium chloride treatment. The Tokomaru soil contained smaller amounts of pH dependent charge, and there was no difference in K+ adsorption between the two K sources. Such a result suggested that where potassium chloride solutions are used to simulate urine K additions to the soil, misleading resutls may occur. In the field, K+ adsorption, and therefore leaching losses due to rainfall, from an application of urine may vary between soil types that differ in their surface charge characteristics. The redistribution of urine K due to leaching, plant uptake and soil adsorption after an application of dairy cow urine to soil was investigated in two experiments with intact cores of four different soil types. In climatic conditions that simulated those of August/September in the Manawatu region up to 80% of the urine moved preferentially beyond the 15 cm depth of soil. This preferential movement of urine occurred too quickly for adsorption reactions to occur between the soil surface and the solutes in the urine, therefore the soil type had little effect on the amount of urine K that was leached. After the urine application, sufficient simulated rain (as 6 events of 5 mm per day applied every second day over a 6-8 week period) was applied to produce 300 mm of drainage. Under free draining conditions the rainfall tended to move preferentially through the macropores of the soil cores bypassing the urine K in the soil, and so the amount leached was equivalent to 3-6% of the urine K applied. Where drainage was impeded, the rainfall moved less preferentially and more K (equivalent to 7-27% of the urine K applied) from the soil micropores was leached. A technique was developed to measure the amount of urine K that was lost preferentially beyond the 15 cm depth of soil in the field situation. This technique was used on seven soil types, four of which were sampled in two different seasons. Results showed that between 0 and 46% of the urine K (and nitrogen) moved beyond the 15 cm depth due to preferential flow. These field results confirmed that preferential flow of urine through soil macropores resulted in the largest loss of K from all the soils studied. In general, the amount of K that was lost was more dependent on the physical conditions of the soil surface (e.g., water repellent or compacted soil surface, irregular surface microtopography) than the soil type or moisture content. Data generated from this study were used to construct a model (the K Loss model) to predict the extent of K losses from three seasonal supply dairy farms. For one of the farms, on a peat soil, it was possible to show that the predictions of soil K losses made by the K Loss model were comparable with actual changes in soil K status. Similar evaluations on the other sites were impossible because the changes of soil K with time were insignificant in comparison with the native soil K contents. Comparisons were made between the losses of K estimated by the K Loss model and the model used by the New Zealand Ministry of Agriculture and Fisheries in their Computerised Fertiliser Advisory Service (CFAS). The CFAS model did not account directly for the loss that occurs due to preferential flow of urine beyond the 15 cm depth of soil, although in some cases this loss can be considerable (e.g., 48 kg K ha-1 yr-1 on the Massey University No. 4 dairy farm), in addition the CFAS model appeared to overestimate the amount of K lost due to transfer of excreta to unproductive areas of the farm (e.g. the transfer loss was overestimated by 39 kg K ha-1 on the No. 4 dairy farm). More accurate predictions of the amounts of K lost, and therefore the amounts of K fertiliser required to replace these losses, can be made easily using information supplied by the farmer trom which the chronometric detail of the cows excretion pattern can be established.